Open source · Free · macOS · Windows · Linux

Blazer

Smart Data Studio

An open-source desktop app for exploring and analysing data with AI. Chat with your files, run SQL instantly, build agentic analysis pipelines — all running locally on your machine.

Download View on GitHub
Engine Status
DuckDB Engine — stable, primary focus
Blazer Engine — work in progress · MLX (Apple Silicon) only

What it does

Everything in one
data workspace.

AI Chat

Ask questions in plain English. The AI generates SQL, runs it, reads the results, and answers you — using the full result table, not just a row count.

DuckDB Console

A full SQL console powered by DuckDB. Query Parquet, CSV, Excel, and JSON files directly from disk. Syntax highlighting, history, and snippet library included.

Agentic Mode

Describe a goal. The agent plans steps, runs queries, reads data, adapts to findings, and delivers a grounded assessment — autonomously.

Result Pane

Every query result opens in a dedicated pane. Collapse, expand, chart, export as CSV, or scroll-highlight from the chat — all from one view.

Schema Explorer

Automatically detects columns, types, and cardinality from loaded files. Column-level statistics and a data profiler built in.

Multi-Provider LLM

Works with Anthropic Claude, OpenAI, and any local Ollama model. Switch models per session. Custom base URL support for proxies and private deployments.

Under the hood

Two engines.
One is ready.

Stable · Primary

DuckDB Engine

The main query engine today. Handles Parquet, CSV, Excel, and JSON natively with blazing-fast in-process analytics.

Works on: macOS, Linux, Windows
Format support: Parquet, CSV/TSV, XLSX, JSON, directories
Focus: All active development, all SQL features
Work in Progress

Blazer Engine

A ground-up columnar engine built on Apple MLX, designed for GPU-accelerated data computation on Apple Silicon Macs.

Requires: Apple Silicon (M1/M2/M3/M4)
Backend: MLX — Apple's machine learning framework, Metal GPU
Status: Core operators scaffolded, not yet production-ready.
The Python and Node.js bindings are available for early experimentation.

SQL Console

Write SQL, see results
in milliseconds.

Blazer DuckDB
AI Chat
Console
History
Snippets
Schema
nyc_trips.parquet
1
2
3
4
5
6
7
8
9
10 
-- Monthly revenue by year/month
SELECT
  year(tpep_pickup_datetime) AS year,
  month(tpep_pickup_datetime) AS month,
  count(*) AS total_trips,
  round(avg(fare_amount), 2) AS avg_fare,
  sum(fare_amount) AS total_revenue
FROM read_parquet('/data/trips/*.parquet')
GROUP BY 1, 2
ORDER BY 1, 2;
Done in 312ms 47,248,845 rows scanned DuckDB 1.1.0
RESULTS 7 rows
# year month total_trips avg_fare total_revenue
0 2015 1 12,748,986 11.91 151,785,085
1 2016 1 10,906,858 12.49 136,193,166
2 2016 2 11,382,049 12.41 141,298,281
3 2016 3 12,210,952 12.80 156,240,088
4 2016 4 11,916,765 12.65 150,746,877

Agentic Mode

Tell it your goal.
It figures out the rest.

Plan Done

Inspect schema & file structure

Monthly revenue trend analysis

Payment type distribution

Data quality check

Synthesize findings & provide final assessment

G
Analyse the NYC taxi dataset. I want to understand revenue trends, payment patterns, and flag any data quality issues.
AI
duckdb query Q3ZP8X result Done 
-- Payment type breakdown
SELECT payment_type,
  count(*) AS trips,
  round(avg(fare_amount),2) AS avg_fare
FROM read_parquet('/data/trips/*.parquet')
GROUP BY 1 ORDER BY 2 DESC;

Credit card dominates at 67.4% of trips with a higher avg fare ($13.20) vs cash at 21.8% ($10.90). Moving to data quality check…

AI

Final Assessment

Revenue is stable at $12.39 avg fare across 47M trips. Credit card is the dominant payment method. Data quality is high — anomalies (<0.6%) are edge cases, not systematic issues.

⏱ 33.6s · 5 queries · 4 steps completed

Open Source

Built in the open.

Blazer Studio is free, open-source, and privacy-first. No telemetry, no accounts, no cloud. Your data never leaves your machine.

Star on GitHub Report an Issue
RustCore engine
Tauri 2Desktop framework
DuckDBSQL engine
MITLicense

Open Source Notices

Built on great open-source work.

Blazer Studio is built on top of open-source libraries. The following are key dependencies, their licenses, and copyright holders. Blazer Studio is not affiliated with, endorsed by, or sponsored by any of these projects or organisations.

Dependency License Copyright
DuckDB Embedded SQL analytics engine, used as a library dependency via the duckdb Rust crate. Not modified or forked. MIT © Stichting DuckDB Foundation
Tauri Desktop application framework (Rust + WebView). MIT / Apache 2.0 © Tauri Programme within the Commons Conservancy
Apple MLX Machine learning framework used by the Blazer engine (WIP). Apple Silicon only. MIT © Apple Inc.
React UI library powering the frontend. MIT © Meta Platforms, Inc.